H2-NewV1: Screenshot-Based AI Coding Assistant

Disclaimer: This tool is designed for educational purposes to understand how cheating tools work. Using this for actual assessments is unethical and may violate terms of service.

H2-NewV1 is an Electron application that captures screenshots, extracts text using OCR, and uses AI APIs (Gemini, Groq, OpenAI) to analyze coding questions and provide solutions.

Features

• Stealth Mode: Transparent, click-through overlay that's invisible to screenshot-based proctoring
• Screenshot Capture: Global shortcuts to capture coding questions
• OCR Text Extraction: Uses Tesseract.js to extract text from screenshots
• AI Analysis: Supports multiple AI providers (Gemini, Groq, OpenAI) with automatic fallback
• Multi-Page Support: Combine multiple screenshots for complex problems
• Drag Mode: Reposition the overlay window as needed
• Always-on-Top: Stays visible even in fullscreen applications

Prerequisites

• Node.js (v14 or later)
• npm or yarn
• API keys from one or more providers (Gemini, Groq, or OpenAI)

Requirements

• Node.js (LTS)
• macOS or Windows (tested on latest versions)
• API keys from Gemini/Groq/OpenAI

Installation

1. Clone or download this project

2. Install dependencies:

   npm install

3. Configure your API keys:

   Copy config.example.json to config.json and add your API keys (at least one):

   cp config.example.json config.json

   Edit config.json with your actual keys:

   {
     "geminiApiKey": "your-gemini-key",
     "groqApiKey": "your-groq-key", 
     "openaiApiKey": "your-openai-key"
   }

Usage

1. Start the application:

   npm start

2. Global Keyboard Shortcuts:
   • Command+Shift+S (macOS) / Ctrl+Shift+S (Windows) - Capture screenshot and get AI solution
   • Command+Shift+A (macOS) / Ctrl+Shift+A (Windows) - Multi-page mode (capture multiple screenshots)
   • Command+Shift+R - Reset/clear current session
   • Command+Shift+O - Toggle overlay visibility
   • Command+Shift+M - Toggle drag mode (reposition window)
   • Command+Shift+Q - Quit application

How It Works

1. The app creates a transparent, always-on-top overlay window
2. Press Command+Shift+S while viewing a coding question
3. Screenshot is captured and text is extracted using OCR (Tesseract.js)
4. Extracted text is sent to your configured AI provider(s) with optimized prompts
5. AI solution appears in the overlay with automatic fallback between providers
6. Use mouse wheel to scroll through the solution

Stealth Features

• Click-through: All mouse clicks pass through to applications below
• Transparent background: No visible window frame
• Screenshot-proof: Won't appear in screen captures
• No taskbar icon: Doesn't show in Alt+Tab or taskbar
• Always-on-top: Works even in fullscreen mode

Technical Details

• Built with Electron for cross-platform support
• Uses screenshot-desktop for screen capture
• OCR powered by Tesseract.js for text extraction
• Multiple AI providers: Gemini, Groq, and OpenAI with automatic fallback
• Marked.js for markdown rendering of responses

Ethical Considerations

This tool demonstrates how easily AI can be used to bypass coding assessments. It exposes fundamental flaws in remote technical interviews that rely solely on LeetCode-style problems.

Using this tool for actual interviews or assessments is:

• Academically dishonest
• Professionally unethical
• Potentially illegal (contract violation)
• Harmful to yourself (getting jobs you're unqualified for)
• Harmful to teams (introducing incompetent members)

This project exists to highlight the inadequacy of current assessment methods and encourage development of better evaluation systems.

License

MIT License - Use at your own risk and ethical discretion.

Credits

Inspired by commercial tools like interviewcoder.co, but made free and open-source to demonstrate the problem with current assessment systems.